Effects of predation risk on behavior, hormone levels, and reproductive success of plateau pikas

Abstract

Predation is a strong selective force that affects animal population dynamics either through direct killing or predation risk effects. Although recent empirical research has shown that the non‐consumptive effect of predator risk on prey dynamics can be as large (or even larger) as direct effects, little is known about the strength of predation risk effects in wild populations or the physiological mechanisms that mediate them. Here, we test both the predator‐sensitive food hypothesis and predation‐stress hypothesis in a single system by investigating activity budgets, stress/sex hormone levels, and demography of plateau pikas (Ochotona curzoniae) and their predators in Haibei Alpine Meadow Ecosystem Research Station of the Chinese Academy of Science. During the study period (2009 and 2010), plateau pikas experienced various predation pressures due to many predators being poisoned in 2010. In the year of high predator numbers, pikas spent more time on vigilant duty and less time foraging and they also showed higher plasma corticosterone levels and lower plasma estradiol and testosterone levels. Reproductive success and male with developed testes also reduced in the year of higher predation risk. In general, our results support both the predator‐sensitive food and predation‐stress hypothesis. Predator‐induced risk affects prey reproduction by changes in feeding patterns and stress physiology.

Introduction

Predator–prey interaction is the central force shaping population and community processes (Krebs et al. , Schmitz ). The conventional view is that direct consumption is the most important way by which predators affect prey population dynamics (Krebs et al. ). Recently, some demographic experiments have indicated that nonlethal predator effects also can influence prey populations (Slos and Stoks ). These nonlethal effects include changes in prey behavior patterns and physiology (Luttbeg and Kerby , Creel et al. , Sheriff et al. , Zanette et al. ).

Apfelbach et al. () claim that prey animals “have developed specific behaviors to facilitate recognition, avoidance and defense against predators” (Apfelbach et al. , p. 1124). There is a great deal of evidence that prey generally increase time spent on vigilance and decrease feeding time, or stay longer in refuges in response to predators or predator cues (Lima and Bednekoff , Creel et al. ). Such predator‐induced behaviors are indispensable for prey species to escape from being killed (Apfelbach et al. ). However, there is a trade‐off between anti‐predator behavior and other fundamental activities. According to the predator‐sensitive food hypothesis, an increase in anti‐predator behavior often reduces the benefit of feeding and has a negative effect on reproductive activity or survival as a result of malnutrition (Sinclair and Arcese ). At present, many experimental and observational results of mammal, birds, larval amphibians, and invertebrates all show that predator‐induced behaviors are strong enough to reduce nutrition, and ultimately come at the cost of survival, growth, body condition, or reproduction (Preisser et al. , Creel et al. , Zanette et al. ).

Predator‐induced stress hypothesis suggests another mechanism by which predator‐induced stress can lead to breeding suppression (Moberg , Boonstra et al. ). Recent findings from laboratory and field studies show that predators or predator odors can cause chronic physiological stress that extends to affect reproduction and survival, and, as a consequence, the prey demography (Sheriff et al. , , Boonstra et al. ). The apparent stress reaction induced by predators is the activation of the hypothalamic–pituitary–adrenal (HPA) axis that causes adrenal glands to secrete epinephrine and glucocorticoids (Sapolsky et al. , Apfelbach et al. , Mateo ). Chronic stress is defined here as long‐time activation of the HPA axis of the body to stressor which is either multiple, frequent exposures, or long‐term constant exposure (Boonstra ). The release of glucocorticoids into the blood improves the level of energy mobilization (Zanette et al. ), which may benefit prey by allowing them to avoid direct killing. However, these stress hormones also inhibit the activity of the digestive and immune systems (see review in Sapolsky ), and the production of sex hormones (Romero ), leading to reduced reproductive output (Sheriff et al. , Boonstra et al. ). A number of laboratory and field studies have shown that predators or predator cues can change prey stress physiology and have negative effects on reproduction (Hayward and Wingfield , Eriksen et al. , Blas et al. , Sheriff et al. , Zanette et al. , Boonstra et al. ).

Plateau pikas (Ochotona curzoniae, also known as black‐lipped pikas) are found in alpine meadows and steppes at 3000–5300 m above sea level on the Qinghai–Xizang plateau, which extends through northern Nepal and Sikkim, India (Schaller ). They are chunky, burrow‐dwelling lagomorphs that do not exhibit sexual dimorphism and do not hibernate (Smith and Foggin , Yin et al. ). Plateau pikas are regarded both as a keystone species contributing to diversity of wildlife on Qinghai–Xizang plateau (Smith and Foggin ), and as pests. In the former role, they are the main food resource for almost all predators, and their abandoned burrows can provide nests for small birds and reptiles (Smith and Foggin ). Meanwhile, their burrowing activity increases the infiltration rate of water and reduces overland flow cross the Qinghai–Xizang plateau providing a critical ecosystem service (Wilson and Smith ). In the latter role, in high densities they degrade rangeland and reduce the forage available for domestic livestock (Fan et al. ). Because of this, widespread poisoning of pikas occurs when populations reach high densities (Lai and Smith ). From 2006 to 2010, we studied the reproductive activities and population dynamic of plateau pikas and their predator at the Haibei Alpine Meadow Ecosystem Research Station (HAMERS) of the Chinese Academy of Science. In spring 2010, under the organization of local Animal Husbandry and Veterinary Station, widespread poisoning for plateau pikas occurred in the Menyuan county of Qinghai Province. The majority of plateau pikas in this region had been killed using poison baits (oats and gophacide mixed together), except in our research area owing to our intervention (about 2000 ha). As a result of the poisoning campaign, the predator population density in 2010 was significantly lower than in 2009 due to the secondary poisoning (see Results of current study, Fig. A). In this study, based on differences of predator density in 2009 and 2010, we test both the predator‐sensitive food hypothesis and predation‐stress hypothesis by investigating behavioral traits and physiological changes of plateau pikas. We predicted that feeding time, plasma estradiol (E₂) and testosterone (TESTO) concentrations, reproductive success (RS), and male with developed testes (DT) would be lower in the year of high predator numbers, but vigilance time and plasma corticosterone (CORT) concentrations would be higher in the year of higher predation risk.

Methods

Study area

This study was conducted at HAMERS in Qinghai Province (China) between April and August, 2009 and 2010. HAMERS is located at the northeast portion of the Qinghai–Tibet Plateau, a large valley in the Qilian Mountains (37°29′–37°45′ N, 101°12′–101°33′ E). The climate is continental monsoon type dominated by the southeastern monsoon and high pressure from Siberia, with an average temperature of −1.7°C and ranging from −37.1° to 27.6°C. Winter is long and severe, and summer is short and cool. Major vegetation types include alpine meadow, alpine shrub, and swamp meadow (Yin et al. ). The study site is located about 5 km east of the HAMERS. The entire research area covers approximately 2000 ha, and no fences or nets surround our study sites. Vegetation type at the study site was Kobresia humilis meadow community, with a variety of sedges, grasses, and forbs that served as pika forage.

Predator population monitoring

In the research area, the plateau pika is the dominant small mammalian herbivore and suffers from high predation risk (Smith and Foggin ). A rich predator community lives in the area, with terrestrial predators such as steppe polecat (Mustela eversmanni), alpine weasels (Mustela altaica), foxes (Vulpes ferrilata, Vulpes vulpes), desert cat (Felis bieti), manul (Felis manul), wolves (Canis lupis), and aerial predators such as upland buzzards (Buteo hemilasius), falcons (Falco cherrug, Falco tinunculus, Falco peregrinus), goshawks (Accipiter gentilis), black kites (Milvus migrans), eagle owl (Bubo bubo), and little owls (Athene noctua—Schaller , Lai and Smith ). The Tibetan plateau is largely a treeless environment, and plateau pikas prefer open areas (Yin et al. ). Therefore, visual observation, combined with binoculars, makes it easy to locate and identify avian and mammalian predators. Dates of investigation were from April to August, which covered the entirety of the breeding season for plateau pikas (Yin et al. ). In the middle of each sampling month, sightings of each species were recorded to assess the relative density of avian and mammalian predators while driving (speed of 20 km/h) a distance of 10 km through the entire study area following a fixed route (the middle of study area). The surveys were taken at 07:00–08:00 and 18:00–19:00 hours per day and not less than 7 days a month (from April to August, 2009: 11, 10, 9, 9, and 8 d, respectively; 2010: 12, 10, 10, 7, and 9 d; respectively).

Sampling plot design

In early April 2009 and 2010, six sampling plots (each is 1.0 ha) were selected on the basis of whether they had established pika populations and similar population densities (about 60 pikas/ha). As another standard for selection of study plots, differences in vegetation (coverage, height, dominant plant species, etc.) and topography (slope, altitude, the distance to road, etc.) had to be as small as possible. To reduce interference between plots, the distance from each other was >1000 m. Plots were divided into five temporary sampling plots and one continuously sampled plot each year. Each temporary plot was only used once to investigate the physiological changes of adult pikas (body mass >130 g) from April to August. The continuously sampled plot was used to investigate population dynamics, behavioral changes, number of juveniles, and reproductive condition from April to August.

Animal trapping

During the reproductive period, all adult females breed, producing three to five sequential litters of three to six young at 3‐week intervals (Dobson et al. , Yin et al. ). Dispersal occurs within a narrow window of time just before initiation of the mating season (in early April). After the mixing that results from dispersal, each newly formed family becomes a cohesive social unit and both males and females fend off intruders from other families, and juvenile pikas remain within their natal family range during the annual breeding season (Dobson et al. ). Thus, marked pikas surviving to the next sampling period can be recaptured in the sampling plot.

In mid‐April, pikas were caught with string snares anchored to the ground near the hole with chopsticks (Dobson et al. ). They were then weighed and labeled with metal ear tag to integrate circular plastic sheet with a unique combination of colors to facilitate recognition in subsequent behavioral observations. Metal ear tag was made by Yin using side wall of cans. The fur on the heads of males and the backs of females was also marked with black hair dye to allow us to determine the sex of individuals at a distance. Each marked pika was released in the original capture location. After about 25–27 days, when large numbers of new young began to move on the ground, all young and adult pikas were trapped, marked, and re‐weighed. Live captured juveniles that weighed <30 g were only marked with hair dye. Each year, the trapping operation was performed five times in the continuously sampled plot (trapping time: April 25–28, May 24–28, June 23–28, July 25–28, August 24–28, 2009; April 27–30, May 26–29, June 24–30, July 26–30, August 26–30, 2010). Following each capture, the weight, capture location, sex, and reproductive condition of each animal were recorded.

After trapping of continuously sampled plot, one temporary plot was randomly selected to sample blood (blood sampling time: 2009 was April 29, May 29, June 29, July 29–30, August 29–30; 2010 was May 1, May 30, July 1, July 31, August 31). Twenty male and 20 female adult pikas on each temporary sampling plots were trapped using string snares. Blood samples (300 μL, less than 0.23% of body weight) were immediately collected from the infra‐orbital sinus. Total handling time from initial capture to completion of blood collection did not exceed 2 min. Blood samples were consistently taken by Yin and stored on ice until centrifugation at 5478 g for 10 min (ice storage time was <10 h) to separate plasma from blood cells, then stored at −20°C. After blood sampling and weighing, pikas were released in the original capture location. All samples were kept on ice during transport to Yangzhou University (they were still frozen upon arrival) and stored at −70°C until subsequent hormone assays. All procedures involved in the handling and care of animals were approved by the China Zoological Society and conform to the guidelines of American Society of Mammalogists.

Reproductive indices of different breeding period

Reproductive indices for pikas were expressed as RS and DT. Since it was difficult to determine relationships of juvenile in the family and adult males have important effects on survival of juveniles during the reproductive season (Yin et al. ), we calculated RS as the total number of juvenile pikas beginning to move on the ground divided by the total number of adult pikas in the same period. Developed testes were indicated if the scrotum was heavy and could be seen easily because it contained fully developed testes; DT (%) was the percentage of adult males with DT from the total number of trapped adult males.

Behavior variables and measurements

Recorded behaviors of plateau pikas included the following: (1) concealing: the animal hid inside the burrow, (2) feeding: gathering or chewing vegetation, and (3) vigilance: sitting in the meadow with neck outstretched, standing with front feet off the ground, or raising head during feeding (see also Yang et al. ).

At the active peaks from 07:00 to 11:00 hours and from 16:00 to 19:00 hours, after most pikas were captured for 5 d, observers measured behaviors according to a focal‐animal sampling rule and a continuous recording rule (Martin and Bateson ). Focal individuals were randomly selected from the colony based on color of ear tag and fur dyeing location. Binoculars were used to identify the focal animal and observe its behavior. During each sampling period, observers sat at a distance over 30 m away from the study area and needed to change observation sites at 2‐hour intervals to ensure every marked adult pika was observed at least one time. Each individual was observed for 15 min. The time of each behavior variables of every 15‐min focal‐animal sampling interval was recorded using pocket observer (Noldus Information Technology, Wageningen, The Netherlands).

Measurement of plasma corticosterone and sex hormone levels

Corticosterone, E₂, and TESTO concentrations in plasma were assayed by rat‐specific enzyme‐linked immunoassay with reagents supplied by Shanghai Yope Biotech Inc. (Shanghai, China), and according to the manufacturer's instructions. These kits had been previously validated for plateau pikas by our research group (Yin ). The prepared plasma samples and the standards were placed in separate plate wells and incubated for 1 h at 37°C. The plate was then washed three times with washing solution; streptavidin–horseradish peroxidase solution was added. After 30 min of incubation at 37°C, the plate was again washed three times and chromogen substrates A and B were added and incubated for 10 min at 37°C. Finally, the enzyme substrate reaction was stopped using a stop solution. After zeroing on a blank well, the optical density of the samples was determined at 450 nm using a Metertech microplate reader (BioTek Instruments, Winooski, Vermont, USA). The sensitivity in the analyses was 1.0 ng/mL for CORT, 1.0 pg/mL for E₂, and 0.1 nmol/L for TESTO. The intra‐assay coefficients of variation for CORT, E₂, and TESTO were 6.1%, 7.7%, and 6.3%, respectively, and inter‐assay coefficients of variation for CORT, E₂, and TESTO were 6.7%, 7.1%, and 7.3%, respectively. All samples were tested twice.

Data analysis

Population density of plateau pika was obtained by the MNA method (minimum number of individuals known to be alive—Buckland ). Mean number of raptors and mammalian predators were calculated by summing all raptors or mammalian number divided by entire survey distance per day, which was a measure of relative predator densities. According the identity of each sampling animals, the time spent in concealing, feeding, and vigilance was calculated by summing the duration of each behavior event in each 15‐min focal‐animal sampling interval using The Observer XT 7.0 (Noldus ). In each sampling period, most plateau pikas were sampled only once. If an animal was sampled more than once during each sampling period, mean time of each variable for this individual was used in ANOVA. The assumption of normality was tested by the Kolmogorov–Smirnov test, and the assumption of homogeneity of variances was tested with Levene's test. If these assumptions were not met, data were logarithmic or square‐root transformed. The relationships between population density and mean percentage of time spent in concealing, feeding, and vigilance, CORT, E₂, and TESTO concentrations in plasma across each sampling periods were evaluated using linear regression analysis. Mean number of sightings of mammalian and raptor predators, body weight of adults, and E₂ and TESTO concentrations in plasma were compared using a two‐way (year × month) ANOVA. CORT concentrations in plasma were compared using a three‐way (sex × year × month) ANOVA. Post hoc Tukey's test was used when significant month effect was detected. Repeated‐measure ANOVAs were conducted to test the effects of sex, month, year, and their interactions on time spent in concealing, feeding, and vigilance. With reproductive indices of plateau pika in 2009 as the expected value and those in 2010 as the observed value, we compared the changes of RS and DT between 2009 and 2010 using χ² statistics. An α level <0.05 was used for all tests. All statistical analyses were carried out using SPSS 16.0 (SPSS Inc. Chicago, Delaware, USA).

Results

Population dynamics

We estimated predator population 67 times in 2009, and 71 times in 2010. Steppe polecat, alpine weasels, wolves, upland buzzards, falcons, saker falcons (F. cherrug), black kites, eagle owl, and little owls were seen during predator population surveys. Of these, the most commonly predators were raptors. Two‐way ANOVA tests indicated that the mean numbers of sightings of predators significantly decreased in 2010 (0.384 ±0.057 km⁻¹) compared with 2009 (0.941 ± 0.086 km⁻¹; mammal: F_1,85 = 4.541, P = 0.038; raptor: F_1,85 = 27.437, P < 0.001; Fig. A). The mean numbers of sightings of raptor predators varied strongly with month (F_4,85 = 3.439, P = 0.023), but no significant month effect and year × month interaction were detected (Appendix S1: Table S1). The mean numbers of sightings of raptors in April are significantly lower than that of July (P = 0.043) and August (P = 0.047).

The plateau pika population displayed different dynamics in the 2 yr (Fig. B). We captured and marked 89 females and 77 males in 2009, and 125 females and 98 males in 2010. Over the 2 years, most juvenile pikas appeared and were moving on the ground in late May and late June. In 2009, no juveniles had appeared on the ground from July to August. Linear regression tests showed that population density was not significantly associated with the behavior, hormone variables across each sampling periods (Appendix S1: Table S2).

Male adult weights significantly increased in 2010 (151.63 ± 3.16 g) compared with 2009 (149.72 ± 2.57 g; F_1,178 = 3.965, P = 0.048), but female adult weights have no significantly difference between 2009 (152.37 ± 4.24 g) and 2010 (153.26 ± 5.06 g; F_1,197 = 0.276, P = 0.609; Fig. ). Two‐way ANOVA tests also show female and male weights varied strongly with month (Fig. B; Appendix S1: Table S1). Female adult weights in July (138.91 ± 1.811 g) and August (133.42 ± 1.41 g) were significantly lower than that of April (153.51 ± 2.38 g), May (156.09 ± 1.39 g), and June (151.95 ± 1.84 g). Post hoc Tukey's test also indicated that female adult weights significantly decreased in August (147.89 ± 3.27 g) compared with April (152.31 ± 2.97 g; P = 0.019). However, a significant year × month interaction in male adult weights was detected (Appendix S1: Table S1), which indicates the variation of male adult weights between 2009 and 2010 is confounded with month effects and depend on the change of month.

Reproductive success of pikas on 24–29 May and 23–30 June was not significantly different between 2009 and 2010 (χ² = 0.297, P = 0.586; χ² = 1.372, P = 0.242, respectively; Table ). However, RS of pikas on 25–30 July and 24–30 August was significantly increased in 2010 compared with 2009 (χ² = 19.222, P < 0.001; χ² = 4.203, P = 0.040 respectively).

Reproductive indices of plateau pika (Ochotona curzoniae) in 2009 and 2010 in Haibei Alpine Meadow Ecosystem Research Station of the Chinese Academy of Science

Period	2009	2010
RS	DT	RS	DT
25–30 April	0 (0/68)	96.88 (31/32)	0 (0/63)	100 (30/30)
24–29 May	1.25 (64/51)	95.65 (22/23)	1.44 (78/54)	100 (24/24)
23–30 June	1.06 (34/32)	12.5 (2/16)	1.56 (56/36)	50.00 (7/14)
25–30 July	0 (0/23)	0 (0/12)	0.71 (20/28)	36.36 (4/11)
24–30 August	0 (0/17)	0 (0/7)	0.27 (6/22)	0 (0/9)

Notes

RS, reproductive success of total adult pikas (number); DT, developed testes ratio of male adult pikas (%). Numbers in parentheses represent sample sizes: The former is the number of juveniles or male adult pikas, and the latter is number of total adult pikas.

Developed testes of adult pikas was higher from April to May, and it sharply decreased from June to August. Developed testes of pikas in 25–30 April and 24–29 May was not significantly different between 2009 and 2010 (χ² = 0.953, P = 0.329; χ² = 1.066, P = 0.302, respectively). However, DT of adult pikas in 23–30 June and 25–30 July significantly increased in 2010 compared with 2009 (χ² = 5.007, P = 0.025; χ² = 6.833, P = 0.009, respectively). In the final trapping period (24–30 August), no males with DT were trapped in the 2 yr (Table ).

Behavior time budget

A total of 318 individuals were observed over the 2 years (2009: from April to August, N = 43, 38, 32, 23, 17, respectively; 2010: from April to August, N = 45, 39, 34, 26, 21, respectively). Feeding and vigilance were the most common behavioral state of surface activity, accounting for 80–97% of average activity budget (Fig. A, B). Feeding time of adult pikas significantly increased in 2010 (10.894 ± 1.494 min) compared with 2009 (8.473 ± 1.529 min; F_1,298 = 20.887, P < 0.001). Feeding time of adults also varied with month (F_4,298 = 7.407, P = 0.007), but no sex effect and interactive effects among year, month, and sex were detected (Appendix S1: Table S3). Feeding time of adults significantly decreased in August compared with May (P = 0.036), June (P = 0.027), and July (P = 0.031).

Vigilance time of adult pikas was significantly lower in 2010 (4.057 ± 1.432 min) than in 2009 (5.901 ± 1.936 min; F_1,298 = 43.367, P < 0.001). Vigilance time of adults also varied from April to August (F_4,298 = 7.036, P = 0.009). Vigilance time of adults was significantly higher in April compared with June (P = 0.016) and July (P = 0.007). Vigilance time of male adults (5.314 ± 1.728 min) was significantly higher than female adults (4.643 ± 1.618 min; F_1,298 = 4.661, P = 0.023), but no interactive effects among year, month, and sex were detected (Appendix S1: Table S3).

Concealing time of adult pikas significantly increased in 2010 (2.268 ± 0.607 min) than in 2009 (1.755 ± 0.547 min; F_1,298 = 40.341, P < 0.001; Fig. C). The concealing time of adult pikas varied with month (F_4,298 = 19.142, P < 0.001). Concealing time significantly increased in August compared with April (P = 0.007), May (P = 0.013), and June (P = 0.024). Concealing time of adult pikas also significantly increased in July compared with April (P = 0.011), May (P = 0.019), but no interactive effects among year, month, and sex were detected (Appendix S1: Table S3).

Plasma corticosterone and sex hormones levels

Blood samples of 309 adult plateau pikas were taken during the two yearly research periods (2009, N = 147; 2010, N = 162). Plasma CORT concentrations of adult pikas significant decreased in 2010 compared with 2009 (F_1,289 = 24.814, P < 0.001). CORT concentrations of female adult pikas decreased 49.53% from 2009 (36.18 ± 8.39 ng/mL, N = 72) to 2010 (17.92 ± 2.42 ng/mL, N = 79) and male adult pikas decreased 53.45% from 2009 (33.41 ± 5.72 ng/mL, N = 75) to 2010 (17.85 ± 1.83 ng/mL, N = 83; Fig. A). However, no month effect, sex effect, and interactive effects among year, month, and sex were detected (Appendix S1: Table S3).

Plasma TESTO concentrations of male adult pikas significantly increased in 2010 (19.947 ± 1.918 nmol/L) compared with 2009 (16.849 ± 1.472 nmol/L; F_1,146 = 9.119, P = 0.027; Fig. ). Two‐way ANOVA test also showed a significant month effect (F_4,146 = 8.019, P = 0.039). Plasma TESTO concentrations significantly decreased in August compared with April (P < 0.001), May (P = 0.003), and June (P = 0.001), but no interaction among year × month groups (Appendix S1: Table S1).

Plasma E₂ concentrations of female adult pikas significantly increased in 2010 (187.876 ± 36.476 pg/mL) compared with 2009 (132.112 ± 15.298 pg/mL; F_1,131 = 12.723, P = 0.023; Fig. ). Plasma E₂ concentrations significantly decreased in August compared with April (P < 0.001), May (P < 0.001), and June (P < 0.001). Plasma E₂ concentrations also decreased in July compared with April (P = 0.006) and June (P = 0.003), but no interaction among year × month groups (Appendix S1: Table S1).

Discussion

Although the predator‐sensitive food hypothesis and the predator‐induced stress hypothesis are not mutually exclusive, to the best of our knowledge, the two hypotheses have not been tested simultaneously in field conditions. In general, our results support both the predator‐sensitive food hypothesis and predator‐induced stress hypothesis. When predator numbers declined in 2010 compared with 2009, plateau pikas decreased vigilance effort and increased feeding time, which increased foraging efficiency and significantly improved male adult weights. This is consistent with predator‐sensitive food hypothesis. Meanwhile, plateau pikas reduced glucocorticoid production and increased plasma estradiol or TESTO concentrations and reproductive output in 2010, which also support predator‐induced stress hypothesis.

More and more research confirms that predators affect prey dynamics through the increased cost of anti‐predator behaviors, even when direct killing is not apparent (Preisser et al. , Schmitz , Zanette et al. ). In many cases, prey respond to high predation risk by reducing their foraging activity, general mobility and by increasing vigilance and their use of safer microhabitats (for a detailed review, see Apfelbach et al. ). Prey still must maintain moderate feeding activity to meet minimum energy demands when exposed to long‐term predation risk (Lima and Bednekoff ). If prey must feed to prevent starvation and successfully evade predators, they have to increase their anti‐predator efforts. Such predator‐induced behavioral changes inevitably increase energy requirements and decrease foraging efficiency, eventually leading to a change in prey demography because of energy deficit (Creel et al. , Zanette et al. ). In our research, the predator numbers were higher in 2009 compared with 2010 (Fig. A). According the predation risk level, the feeding time and male adult weights significantly decreased in the year of higher predation risk (Figs.  and ), which probably decreased foraging efficiency and increased nutritional stress. These results support the predator‐sensitive food hypothesis in a mammal under field conditions. Our results also showed that adult weights significantly decreased during the research time. The variation of female adult weights might result of the effect of gestation. Nevertheless, the reduction of male adult weights might result of the cost of paternal care and mating competition (Yin et al. ). Males emit alarm calls when deter a predator and spent more time on the vigilance task at April (Fig. ), which probably increased nutritional stress and reduced energy storage.

The adjustment in anti‐predator responses to cope with predation risk levels might be driven by stress response (Monclús et al. ). The elevations of glucocorticoids resulting from predators or predator cues enhance the concentration of blood sugar and metabolic rate, which can be used in the typical “fight or flight” response (Sapolsky et al. ). Recent field researches suggest that predator or predator cues can impact the stress physiology of prey (Mateo , Sheriff et al. , Clinchy et al. ). According to the predator‐induced stress hypothesis, elevation of glucocorticoids in response to stressors not only reduces energy storage, but also suppresses the activity of the digestive system and immune organs (see review in Sapolsky ), finally leading to breeding suppression, especially in chronic exposure to higher glucocorticoid concentrations (Blas et al. ). A large number of laboratory studies confirm that predator‐induced chronic stress can reduce the amount of sex hormones and affect prey animals’ reproduction (Romero , Reeder and Kramer , Blas et al. , Mateo ). Recently, field studies on snowshoe hares (Lepus americanus) and yellow‐bellied marmots (Marmota flaviventris) show that predation risk elevates fecal glucocorticoid metabolite concentrations and significantly inhibits reproductive output (Preisser et al. , Monclús et al. ). In our study, plasma CORT concentrations of adult pikas significantly increased, but plasma sex hormone concentrations significantly decreased in the year of higher predation risk (Fig. ). In addition, RS of adult pikas in 25–30 July and 24–30 August decreased in the year of higher predation risk (Table ). These results are consistent with the predator‐induced stress hypothesis.

In contrast, field experiments on elk suggest that exposure to predators does not change fecal glucocorticoid concentrations, but does affect their foraging patterns and habitat selection and extends to affecting breeding (Creel et al. , Christlanson and Creel ). Boonstra () claims that predator‐induced stress response is “an adaptive trait selected for under certain life histories” (Boonstra , p. 20). When in danger, elk usually run away and shift foraging areas from preferred grassland to poorer quality but less risky forest (Creel et al. ), which perhaps reduce stress responses. Hence, the change in nutrition associating with predation risk is possible the main factor impacting elk reproduction (Creel et al. , Clinchy et al. ). As burrow‐dwelling species, interconnected burrowing system is an important shelter for plateau pikas to copulating, caring for young, and avoiding predator hunting and killing. However, in high‐risk environment, staying in burrow not ensured to reduce predation risk because some mammalian predator (steppe polecat, alpine weasels, etc.) can get into burrow to hunt them. Thus, if higher danger is encountered more frequently, plateau pikas have to decrease the concealing time (staying in burrow) and increase the vigilance time to reduce predation risk, which most likely increased stress response and extend to cause a decline in reproduction. In our results, the concealing time significantly decreased, but vigilance time significantly increased in the year of higher predation risk, which supports the predator‐induced stress hypothesis.

Factors such as variable habitat quality, food supply, or population density could explain the difference in behavior and hormones (Sheriff et al. ). However, there were no obvious visual differences on food resource (plant cover and height) and habitat quality between the 2 years, and monthly rainfall and mean temperatures were also similar (Fig. , data were obtained from the meteorological station of HAMERS situated 6 km from the study area). Although high population density is associated with increased glucocorticoid secretion in mammals (Creel et al. ), the difference of population density (Fig. B) between the two years is unlikely to be substantial enough to change the behavior and hormone. Linear regression test on the relationships between population density and behavior, and hormone variables across each sampling periods show no statistical significance (Appendix S1: Table S2). However, predator numbers were reduced 59.24% in 2010 compared with 2009 due to the pika poisoning campaign (Fig. A). Predator numbers decreasing is bound to reduce direct killing and non‐consumptive effect on plateau pikas population dynamic, especially for non‐consumptive effect. For example, research on free‐living song sparrows (Melospiza melodia) has shown that non‐consumptive effect of predators causes a 40% reduction in the number of offspring produced per year (Zanette et al. ). Although we could not isolate the non‐consumptive effect of predation risk on RS, our finding showed that predation risk clearly changed the behavior and physiology of plateau pikas, particularly in reproductive activity. Thus, plateau pika population dynamics in the 2 yr might have resulted mainly from the change in the predator population. Predation is the predominant organizing process in prey population dynamics, acting both directly through mortality and indirectly through a non‐consumptive risk effect.

Acknowledgments

We thank two anonymous reviewers for constructive comments and the numerous field assistants who helped with the collection of data. This research was supported by National Basic Research Program of China (973 Program, 2007CB109102) and the National Natural Science Foundation of China (No. 31272320).